Dual CPU F@H run as service..

DonP · Post by **DonP** » Wed Sep 10, 2003 3:51 pm

Hi,
I've got a query.. I'm trying to run the console version of F@H on a dual CPU box.. I'd really like to run it as a service (or as two services I guess).. but I've hit a snag.. the firedaemon method requires Administrator priveleges and I think I've only got Power User status.
I just found this which I'll try tomorrow but I'd really like to capture the text output for debugging... I'm guessing this second method won't save the text output to a file.

So, any ideas on how I could either install FireDaemon without admin rights, or to capture output of f@h directly installed as a service?

Thanks,
DonP.

Post by **Rusty075** » Wed Sep 10, 2003 4:06 pm

A better solution would be this: OverClockers Australia F@H service guide

As for capturing the output, you don't have to do anything. F@H automatically outputs all of its text into a log file. Look in the directory that your F@H is running in for a file named FAHLog.txt

DonP · Post by **DonP** » Thu Sep 11, 2003 5:13 pm

Big Thanks Rusty....
unfortunately I've hit a snag - it seems I can't edit the "Services" part of the registry.. so I've given up on running it as a service and I'll just run it from a script and/or "Startup" (just run it when a user logs in).

But now.. and this is the bit which makes me absolutely mad.. I can't get two instances to run.
From what I see I need to create two separate directories with one copy of the fah3console.exe in each. Then I run "fah3console -config" in one dir to create the config file. I use machineid=1 for this.. then I copy it to the second fah dir and in that config file I change machineid to 2.
Great so far, just like it says in the FAQ. Note I didn't just copy an installed directory over, I haven't done what is says not to do here. So, I've just two dir, each with a fah3console.exe, each with a client.cfg, differing by MachineID.

Now if I run one ("C:\dir1\> fah3console -local"), it works. Now if I start the second ("C:\dir2\> fah3console -local") the client.cfg of that instance ("C:\dir2\client.cfg") gets overwritten with some simple three line (incorrect) config. Once I replace C:\dir2\client.cfg and run it ("C:\dir2\> fah3console -local") then the config of the first one gets overwritten.

ARGH! I've browsed the web a bit (I gave up after an hour of fiddling) and it seems to me that the Stanford FAQs, "fah3console --help" and the few posts I've found regarding client.cfg are all about as useful as a serious heart attack.

Oh.. and one other problem.. I looked at the logs and it seems I have about 10 WUs (from running the graphical version) ready for upload but they cannot be uploaded... I'm connecting through a proxy server but I know that it's worked in the past.. and now after fiddling with the console clients it doesn't. I don't think this problem is specific to this dual CPU box though - another single CPU box hasn't been able to send WU either. I've tried three proxies. I checked the status of the server and supposedly they work.. but when I telnet to them (on port 8080) it connects and then just disconnects.

This whole F@H things is really infuriating with the really poor documentation and debugging features ("-verbosity 9" is useless).

Any ideas? am I missing something obvious? I'm pretty sure I've tried all combinations but maybe I've missed something.

BTW.. if I'm behind a proxy running two separate (say single CPU machines for now) F@H hosts then I don't need to fiddle machineid for each do I?

Thanks and sorry for the general grumpy tone,
DonP.

PS One last Q.. if I copy the "work" directory from the PCs which cannot return the results to a computer at home, which repeatedly has successfully returned results then will it work? Or do the files in "work" have something specific to the machine the were computed on?

DonP · Post by **DonP** » Thu Sep 11, 2003 7:29 pm

I've been doing more browsing and I can't see a mention of the 2cpu problems I;ve been having so it must be specific to my setup - I;ll dig around more next week.

But if you have any idea on why complete WU aren;t being returned then I'd still like to know.. also the question about returning WUs on a different machine to which they were computed.

Thanks!

WarpedPlatter · Post by **WarpedPlatter** » Fri Sep 12, 2003 7:14 am

DonP wrote: PS One last Q.. if I copy the "work" directory from the PCs which cannot return the results to a computer at home, which repeatedly has successfully returned results then will it work? Or do the files in "work" have something specific to the machine the were computed on?

On a few occasions I've successfully started a work unit on one computer and moved the work directory to another to finish it.

Lockheed · Post by **Lockheed** » Fri Sep 12, 2003 10:37 am

Post removed

DonP · Post by **DonP** » Sun Sep 14, 2003 2:50 pm

Thanks guys for all the advice - I got the 2 CPU issue resolved - I think there may be an un-initialised variable somewhere or something - started to work fine after a reboot.. let's see how long it stays fine

About the uploading of results - I still can't do it... I've tried three different proxies (2*squid, 1*tinyproxy), one for which I could access the logs to.. the connection goes through but then no data is sent either way.. I also tried sending it through port 443.. hopefully the httpS port won;t cache and just create a clear tunnel.. but no luck.
I don't get it... and I'm a bit fed up of this hacking - it should really be a trivial thing.
It's a shame it's borked cos I have quite a few PCs behind a firewall and if I can't automatically upload results then I might as well start burning my CPUs to cure cancer rather than folding.
I've searched around for similar problems - I found a few threads and their resolution was just that it started to work magically.. I suspect the same will happen in my case since it _did_ actually work right at the start.

Anyway.. I'm going to carry on folding.. can;t wait for the extremeOC stats to return.. looks like we're 31st.. what can we do about SLO-TEch?

Regards,
DonP.

PS It looks like you CAN move finished WUs from the work directory on one machine to another.. and they are successfully uploaded.
EDIT:
PPS Actually... just tried shifting some more complete WUs again and it didn;t successfully transplant.. hmm

PPPS This isn;t the problem I'm having - I've looked at squid.conf and there is no limit on the request size.

Post by **Rusty075** » Sun Sep 14, 2003 7:53 pm

DonP, I'm sure you've tried this, but just to be sure:

These machine do have WWW access through IE, right?

And if so, have you tried the "Use Interent Explorer settings" right?

Silly suggestions, I know, but I'm trying to help

DonP · Post by **DonP** » Mon Sep 15, 2003 12:21 pm

Rusty075 wrote:..And if so, have you tried the "Use Interent Explorer settings" right?

I have, thanks Rusty, but it doesn't help.
I did get a bit of a break today.. I noticed that a few WU recently got through.. and they were smaller than 1M.. so it looks like I am actually having this problem.
I didn't think it was that problem at first because on the dual CPU box where I was doing most of my testing on, it didn;t give me an error 413 (even with verbosity 9). I got a 413 today on another box. I also checked the config file of squid and there was no mention of request_body_max_size and the default should be no limit (I think).
So tomorrow I'm going to come up with a good reason for the squid admin to up the limit.

Thanks and sorry for my angry tone - I'm just annoyed that this isn't click-and-go.

DonP.